Identification of potential pharmacogenomic markers of clinical efficacy of 5-fluorouracil in colorectal cancer

Abstract

Although adjuvant chemotherapy has significantly increased overall survival in resected Stage III colorectal cancer, disease recurrence is still high (30–40%). 20–25% of Stage II patients also develop recurrent disease. Thus, high-risk patients may benefit from chemotherapy. As patient response to standard chemotherapy varies, the study of molecular differences in the expression of pharmacologically relevant genes may help clinicians to understand variability and tailor therapy. The expression of 5-fluorouracil (5-FU) pathway genes in tumors from 53 Stages II-III colorectal cancer patients who underwent 5-FU adjuvant chemotherapy was investigated by reverse transcription quantitative real-time polymerase chain reaction. Patients were dichotomized into high- and low-mRNA expression level groups using median values of gene mRNA levels. Then, a threshold analysis to identify a cut-off distinguishing recurrent- or nonrecurrent-disease was used. A high degree of interpatient variation in relative tumor expression of study genes was observed. Multiple gene correlations were found, which suggest possible coregulation mechanisms. No statistically significant relationship between experimental data and baseline clinical/pathological characteristics or clinical outcome was observed using gene expression median values. Threshold analysis indicated significant inverse relationships between deoxyuridine triphosphatase (DUT), ferrodoxin reductase (FDXR) or tumor protein p53 (TP53) and disease-free survival (DFS) in the entire case series and between DUT or NM23-H1 and DFS in Stage III patients: higher gene expression was associated with shorter DFS. This study provides data on relationships between expression of 5-FU pathway genes and clinical outcome of colorectal cancer patients undergoing 5-FU adjuvant chemotherapy and underscores the predictive role of specific genes. Validation in an independent case series is warranted.

Colorectal cancer is the fourth most common cancer in terms of incidence and the second in terms of mortality in the Western world.1 Although improvements have been made in the medical treatment of colorectal cancer patients, both in advanced and in the adjuvant setting, therapeutic outcome remains unsatisfactory after chemotherapy because of inherent or acquired drug resistance.2 Despite adjuvant chemotherapy, a relevant percentage of patients (30–40%) with regional lymph node involvement (Stage III) develop disease recurrence within 5 years from surgery.3, 4 Also, the benefit of adjuvant chemotherapy in Stage II patients remains controversial.5, 6 An early meta-analysis of prospective clinical trials in Stage II patients showed no survival benefit for those who were given chemotherapy vs. untreated control patients.7 In contrast, evidence for an improvement in survival due to adjuvant chemotherapy in Stage II patients has been more recently reported by an analysis of individual patient data from 18 trials with about 6,900 patients.8 However, 20–25% of Stage II colorectal cancer patients still develop recurrent disease.5, 9

Current results in the adjuvant setting of colorectal cancer have been obtained using fluoropyrimidine-based chemotherapy regimens with or without oxaliplatin.4 These chemotherapy regimens are administered to all Stages II–III patients due to the impossibility of selecting them on the basis of the knowledge of molecular determinants predictive of treatment outcome. This information may help to characterize colorectal cancer patients who should receive specific chemotherapy, which could improve the therapeutic index, their survival and quality of life.

Intense research on the mechanisms underlying the drug resistant phenotype has led to the identification of individual target genes whose expression levels or polymorphisms can predict response to specific anticancer drugs.10 However, individual markers are not sufficient to predict response with the degree of sensitivity and specificity that is required for clinical practice.11 Gene expression profiling of pathways associated with metabolism and the action of chemotherapeutic drugs may help to define predictive patterns of clinical drug effects in colorectal cancer more accurately,12, 13 as shown for breast cancer.14

Retrospective studies indicate that metastatic colorectal cancer patients displaying increased tumor enzymatic activity, increased mRNA or protein expression of thymidylate synthase (TS), the main target of 5-fluorouracil (5-FU)15–18 or altered expression of enzymes of 5-FU metabolism such as thymidine phosphorylase (TP),16 orotate phosphoribosyltransferase (OPRT),19 uridine monophosphate kinase (UMPK)20 and dihydropyrimidine dehydrogenase (DPD)16 are usually not responsive to 5-FU-based chemotherapy. Results from a prospective clinical study in metastatic colorectal cancer patients in which chemotherapy selection (5-FU or alternative drugs, i.e., combined oxaliplatin, OHP and irinotecan, CPT-11) was based on mRNA expression levels of only two such genes [thymidylate synthase (TYMS) and DPD] have been disappointing.21

Thus, the current research is focused on the analysis of a higher number of potential molecular markers related to 5-FU pharmacological effects.22, 23 However, to date, no robust evidence for predictive value of molecular markers of 5-FU treatment efficacy has been provided.

We quantitatively analyzed the expression of 26 genes involved in the metabolism and mechanism of action of 5-FU in tumor samples obtained at surgery from patients with Stage II or Stage III colorectal cancer who underwent 5-FU adjuvant chemotherapy to establish the possible correlations between candidate gene expression levels and clinical outcome and to verify whether these genes represent a predictive determinant of response to 5-FU adjuvant chemotherapy in colorectal cancer.

Material and Methods

Patients and sample collection

Fifty-three tumor tissues (collected from 1993 to 2002) from Stages II-III colorectal cancer patients (AJCC Stage classification) who underwent 5-FU adjuvant chemotherapy were selected from our tumor tissue bank. All patients received folinic acid and 5-FU according to the schedule described in the pooled analysis by the IMPACT investigators.24 Informed consent was obtained from patients regarding the use of the specimens and clinical/pathological data for research purposes and all procedures followed the guidelines established by the local ethical committee. Primary tumor explants obtained from colorectal cancer patients at surgery were frozen in liquid nitrogen until molecular analysis. Immediately after resection, the tumor sample was divided into two equal portions after washing and removal of necrotic tissues. One portion was fresh frozen in liquid nitrogen until the time of RNA extraction, and the other portion was embedded in paraffin to confirm histologically that it was not significantly contaminated by normal tissues, necrotic tissues and lymphocytes.

Baseline clinical and pathological characteristics of patients are summarized in Table 1. Complete follow-up was available for all these patients for at least 5 years (mean follow-up, 7.4 years).

Table 1. Main clinical and pathological characteristics of patients

Disease recurrence occurred in 17 of 53 patients; the median time between surgery and this event was 16.3 months (range 4.1–40.0). The median follow-up for the 39 patients alive at the moment of the analysis was 95.2 months (range 46+ to 165.7+).

Gene expression analysis

The 26 candidate 5-FU pathway genes studied were selected on the basis of their potential role in 5-FU activity, either at the metabolic (activation and detoxification) level or at the target and the post-target levels (see additional Supporting Information). Total RNA was isolated using a Trizol RNA isolation kit with glass-fiber filter purification methodology (RiboPure kit Ambion, Austin, TX). Concentration and purity of the RNA were verified by Gene Quant II spectrophotometer (Pharmacia Biotech, Cambridge, UK). The absorbance ratio at 260/280 nm of all the samples ranged from 1.8 to 2.0 indicating they were all free from contaminants. The RNA integrity was been assayed on a 0.8% agarose gel stained with ethidium bromide. The ratio of the intensities of 28 S to 18 S was ∼2:1 through all samples. This control enabled us to consider all RNA samples suitable for reverse transcription (RT) and real-time polymerase chain reaction (PCR) assays.

cDNA was generated from 10 μg of total RNA using random primers and the M-MLV reverse transcriptase RNase H minus (Promega Corp., Madison, WI) according to the manufacturer's protocol.

Real-time PCR analysis was performed with the ABI PRISM 7900HT Fast Sequence Detection System (Applied Biosystems, Foster City, CA).

Predesigned and validated gene-specific probe-based TaqMan gene expression assays from Applied Biosystems (Foster City, CA) were used for the target study genes. Every set contained gene-specific forward and reverse primers as well as fluorescence-labeled probes. Reactions were performed using Taqman Fast Universal PCR Master Mix No AmpErase UNG and two to three replicates for each reaction were plated into 96-well plates. The amplification profile was one cycle of denaturation for 20 sec at 95°C followed by 40 cycles with 1 sec of denaturation at 95°C and annealing, extension for 20 sec at 60°C.

To standardize all the samples, the housekeeping gene glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used as an endogenous reference gene (TaqMan endogenous control concentration-limited primer; Applied Biosystems, Foster City, CA). Human reference total RNA (Stratagene, La Jolla, CA) was used as a calibrator sample to compare the expression of the different genes. The mRNA target gene expression levels were normalized to the endogenous reference and expressed in relationship to the calibrator as 2^−ΔΔCT (comparative CT method). A validation experiment was performed to demonstrate that the efficiencies of target genes and reference gene amplifications were approximately equal, using a standard curve method with several dilutions of the cDNA calibrator sample.

Statistical analysis

The correlation of expression values of the 26 5-FU pathway genes with clinical and pathological characteristics of patients was analyzed using the Spearman test (age), the Wilcoxon RankSum test (sex and histotype) and the Kruskal–Wallis test (tumor site, stage and grading).

Gene correlations were evaluated using the Spearman rank correlation with a value >0.600 deemed to be biologically significant. Additionally, an unweighted pair grouping method, unsupervised hierarchical clustering was used to examine gene expression relationships and patient groupings. The software used was the Cluster-Tree View version 1.6, available at http://rana.lbl.gov/EisenSoftware.htm.

As a first level of analysis, patients were dichotomized into high- and low-mRNA level groups using the median value of mRNA levels as a cut-off value. The accuracy of gene expression in predicting disease-free survival (DFS) was also evaluated by receiver operating characteristic (ROC) analysis. The ROC curve provides a visual description of the trade-off between false-positive and true-positive rates for all possible threshold values. The optimal cut-off value for differentiation of patients with presence or absence of disease recurrence following 5-FU adjuvant chemotherapy was defined by the point of the ROC curve with minimum distance from the 0% false-positive rate to the 100% true-positive rate.

Univariate and multivariate Cox proportional hazard regression was used to identify molecular markers correlated with patient DFS. The t-test was used to select molecular markers that distinguished between patients with recurrent and nonrecurrent disease. Molecular markers that were found by both Cox model and t-test were selected to build a signature for predicting outcome.

To define whether disease stage affected the relationship between gene expression and outcome, an independent analysis for subgroups of Stages II and III patients was also carried out.

Kaplan–Meier survival plots and the log-rank test were used to assess the differences in DFS or overall survival (OS) between patients' low- and high-gene expression levels. DFS time was calculated from the date of diagnosis until disease progression. OS time was calculated from the date of diagnosis until death or last follow-up. Analyses were carried out using the SPSS version 13 software. p values <0.05 were considered significant.

Results

A high degree of interpatient variability in relative tumor expression of 5-FU pathway genes was observed (Fig. 1).

Figure 1. Interpatient variability in relative expression (vs. GAPDH) of 5-FU pathway genes. Long bars, range; large rectangle, middle 50% of patients samples; small square, mean value; bar, median values; and the red circles show the largest and the smallest range of relative expression (DPYS and RRM1).

Multiple gene correlations were found, which suggest possible coregulation mechanisms (p < 0.001) (Table 2). 5′,3′-Nucleotidase cytosolic (NT5C), methylenetetrahydrofolate reductase (NAD(P)H) (MTHFR), nuclear factor of kappa light polypeptide gene enhancer in B cells 1 (NFKB1), uridine-cytidine kinase 2 (alias uridine monophosphate kinase, UMPK), tumor protein p53 (TP53), ferredoxin reductase (FDXR) and folylpolyglutamate synthase (FPGS) each showed correlations with 20–16 other genes. Dihydropyrimidine dehydrogenase (DPYD), dihydrofolate reductase (DHFR), uridine monophosphate synthetase (UMPS), B-cell CLL/lymphoma 2 (BCL2), dihydropyrimidinase (DPYS), endothelial cell growth factor 1 (ECGF1, alias TP), protein (NM23A) expressed in non-metastatic cells 1 (NM23-H1), ribonucleotide reductase M1 (RRM1), uracil-DNA glycosylase (UNG), uridine phosphorylase (UPP1) and β-ureidopropionase (UPB1) each showed correlations with 15–10 other genes. The remaining eight genes were correlated with fewer than ten other genes (from eight to one).

Table 2. Spearman rank correlation of 5-FU pathway genes

The unsupervised hierarchical clustering algorithm allowed us to cluster the 53 tumors on the basis of similarities of 26 study genes. The length and subdivision of the branches of the dendrograms of Figure 2 display the relatedness of the colorectal cancers (top) and the expression of the genes (left). From this analysis, tumors could be divided into two groups (n = 17, Group 1 and n = 36, Group 2) on the basis of gene expression. Mean gene expression values were generally lower in the first group compared with the other. However, these differences were not predictive of disease relapse: there was no statistically significant difference between patients with disease recurrence in the first group (23.5%) compared with those in the second (36.1%) (p = 0.530).

Figure 2. Hierarchical clustering. Unweighted pair grouping method with arithmetic mean hierarchical clustering using the relative expression data for the 26 genes of the 53 patients was performed, yielding at least two groups of genes (left; specific genes, right) and two groups of patients (across the top). Disease recurrence status for each patient is shown in correspondence with the patient number (white indicates patients who continue to be disease free; black indicates patients showing disease recurrence).

The unsupervised hierarchical clustering method also revealed two main gene clusters (Fig. 2). Cluster 1 had 11 genes with a similarity score of 0.220; it had two subclusters: one included protein (NM23A) expressed in non-metastatic cells 2, deoxyuridine triphosphatase (DUT) and TYMS (similarity score of 0.393) and the other was constituted by further two subclusters: γ-glutamyl hydrolase (GGH) and FDXR (similarity score of 0.532) and UNG, RRM1, NM23H1, UMPS, DHFR and UMPK (similarity score of 0.587), respectively.

Cluster 2 contained the remaining 15 genes with a similarity score of 0.10. Two main subclusters were constituted by three genes [ribonucleotide reductase M2 (RRM2), E2F transcription factor 1 (E2F1) and thymidine kinase 1, soluble (TK1), similarity score of 0.534] and 12 genes (similarity score of 0.409), respectively. The latter subcluster contained two other subclusters represented by a 7-gene subcluster (NT5C, MTHFR, NFKB1, ECGF1, UPP1, DPYS and TP53, similarity score of 0.563) and a 5-gene subcluster (FPGS, DPYD, fas ligand (FASLG), BCL2 and UPB1, similarity score of 0.502), respectively.

Correlations between experimental data and baseline clinical–pathological characteristics indicate a significant difference in the expression of TK1 according to sex (p = 0.034). UPP1 also varied according to age (p = 0.048). RRM2 and TK1 varied significantly in relation to tumor site (p = 0.0021 and p = 0.020, respectively) and BCL2 expression levels varied significantly as a function of grading (p = 0.039). No significant difference in gene expression in relation to disease stage was observed.

Relationships between clinical outcome (survival parameters) and gene expression levels or clinical–pathological characteristics were also studied.

When Stages II and III patients were analyzed together, the segregation of gene expression levels according to the median values showed no statistically significant associations by univariate analysis of DFS or OS. Only a trend between DFS and DUT expression was observed and no significant relationship was observed between gene expression and OS. A higher DUT expression level was associated with a shorter DFS (p = 0.060). Relationships between DFS and OS and clinico–pathological features showed significant correlations only with disease stage (p = 0.014 and p = 0.018, respectively). Multivariate analysis confirmed stage as predictor of DFS (p = 0.014) (data not shown).

DFS was also evaluated on the basis of a threshold analysis able to distinguish between patients with recurrent and nonrecurrent disease in relation to tumor gene expression. In the entire patient population, area under the curve (AUC) values >0.60 were observed for DUT, FDXR and TP53 mRNA expression (Fig. 3). In particular, the expression value that best differentiated between recurrent and nonrecurrent tumors was 10.07 for the DUT gene; those that best distinguished recurrent tumors from nonrecurrent were 6.52 and 9.92 for FDXR and TP53, respectively. Patients whose tumors expressed levels of these genes equal or below these values had a prolonged DFS (Fig. 3).

Figure 3. Correlations between gene expression of DUT, TP53, FDXR and DFS in all patients (Stages II-III, n = 53). (a) Receiver operating characteristic (ROC) curves. (b) Kaplan–Meier curves indicating probability of disease relapse for patients with gene expression above or equal/below the cut-off distinguishing “recurrent” and “nonrecurrent” patients.

The univariate analysis of Stage II patients (n = 27), differentiated on the basis of the median values or of cut-off values obtained by the threshold analysis, showed no relationships between gene expression and survival parameters.

When Stage III patients (n = 26) were differentiated on the basis of median gene expression values, an inverse relationship was observed between DFS and NM23-H1 expression (p = 0.044) and a trend between DFS and DUT (p = 0.063) according to univariate analysis. In both the cases, higher gene expression levels were associated with a shorter DFS (data not shown). An inverse relationship between OS and FDXR mRNA expression levels was observed with patients having a higher gene expression level of FDXR having a shorter OS (p = 0.046) (data not shown).

The threshold analysis gave an AUC >0.6 for seven genes (i.e., DUT, FDXR, NM23-H1, RRM1, TYMS, UPB1 and TP53). However, cut-offs able to significantly differentiate recurrent from nonrecurrent disease were identified only for NM23-H1 and DUT (Fig. 4). A gene expression value of 4.75 was the best cut-off for NM23-H1 in Stage III patients, whereas 10.07 was confirmed as the best cut-off value for DUT as observed in the entire case series. By using the best cut-off values for TYMS and RRM1 (1.59 and 4.16, respectively), a trend between high expression of these genes and reduced DFS (p = 0.073 and 0.064) was observed (Fig. 4).

Figure 4. Correlations between gene expression of DUT, NME23-H1, RRM1, TYMS and DFS in Stage III CRC patients (n = 26). (a) Receiver operating characteristic (ROC) curves. (b) Kaplan–Meier curves indicating probability of disease relapse for patients with gene expression above or equal/below the cut-off distinguishing “recurrent” and “nonrecurrent” patients.

Discussion

About 50% of Stage III colorectal cancer patients experience disease recurrence and die of their disease after potentially curative surgery.3, 4 Today, it is not possible to predict which patients will be cured after surgery from those who are not. Thus, adjuvant chemotherapy is a routine treatment for Stage III colorectal cancer patients, with only about 10–15% improvement in OS.4, 8

A significant proportion of Stage II colorectal cancer patients (20–25%) will also experience recurrent disease and die from it.5, 9 However, the administration of adjuvant chemotherapy in these patients is still controversial. Thus, there is a strong need for prognostic factors, which can identify patients who are likely to experience relapse and for predictive factors able to identify patients who are more likely to benefit from adjuvant chemotherapy, as its therapeutic efficacy varies greatly in patients with the same clinical and pathological characteristics.

The objective of our study was to establish whether candidate gene expression levels can predict response to 5-FU adjuvant chemotherapy in colorectal cancer patients.

Although the evaluation of single genes or a small panel of genes may be useful for predicting benefit from chemotherapy, it is clear that a complex metabolic and action pathway such as that of 5-FU which is used to treat a complex disease such as colorectal cancer requires a polygenic approach to predict outcome. Thus, we analyzed gene expression of the 5-FU pathway in fresh-frozen tumor samples from 53 colorectal cancer patients using RT real-time PCR.

We tried to assess coexpression or coregulation of the 5-FU pathway genes studied. The analysis of the 26 genes studied reveals two main gene clusters and many of the genes that were closely linked in the hierarchical clustering were also correlated in the Spearman rank results showing similar gene groupings.

When gene expression levels are segregated on the basis of the median values, patients whose tumors had a higher expression level of DUT showed a tendency to have a shorter DFS (p = 0.060) in the entire case series (Stages II and III patients). Threshold analysis indicated relationships between DUT, TP53 and FDXR expression and DFS.

Deoxyuridine triphosphate nucleotidohydrolase (dUTPase), the enzyme codified by DUT, is the key regulator of deoxyuridine triphosphate (dUTP) pools. It is responsible for the hydrolysis of dUTP to deoxyuridine monophosphate (dUMP) and pyrophosphate, simultaneously providing a substrate for TS and eliminating dUTP from the DNA biosynthetic pathway. The expression of this enzyme may be an important determinant of cytotoxicity mediated by fluoropyrimidine-induced TS inhibition both in human colorectal tumor cell lines25, 26 and in colorectal tumor specimens from patients.27 Increased dUTPase activity and expression are associated with fluoropyrimidine resistance in vitro and in the clinic, and our results confirm these findings.

The TP53 tumor suppressor gene acts to maintain the integrity of DNA after damage mainly by transcriptional activation of several genes28 and is frequently mutated in colorectal cancer.29 It inhibits entry into S phase and induces cell cycle arrest by transcriptionally upregulating the G1 cyclin-dependent kinase inhibitor p21^WAF1 and G2-M checkpoint genes, respectively. TP53 is also involved in the induction of proapoptotic genes such as FASLG and repression of anti-apoptotic genes such as Bcl-2 and may modulate TS activity.30 Mutations in TP53 or disruption through homologous recombination of TP53 in colorectal cancer cells are related to resistance to 5-FU and lack of sensitization to 5-FU-induced Fas-mediated apoptosis.31, 32

TP53 has been investigated as a predictive factor of colorectal cancer response to therapy with conflicting results. The methods used to assess the mutational status of TP53 greatly vary among studies as do study designs, especially in terms of patient selection, making it difficult to arrive at conclusions of its predictive value.33 Our results show that low levels of TP53 mRNA were associated with prolonged DFS. This finding is in agreement with results from other studies in Stage III colorectal cancer patients treated with adjuvant chemotherapy.34, 35 In patients whose tumors had a normal TP53 status as assessed by immunohistochemistry (p53 protein overexpression) or sequence analysis, adjuvant chemotherapy provided a significant survival advantage, whereas in patients whose tumors overexpressed p53 protein or with mutant TP53, adjuvant treatment did not show benefit.34, 35 However, it has been recently shown that TP53 status does not consistently explain the variance in responses of fluoropyrimidine-treated colorectal cancer cells36 and its role as a marker of therapeutic activity in colorectal cancer requires additional investigation.

FDXR codifies for the protein ferredoxin reductase that transfers electrons from NADPH to cytochrome P450 via ferredoxin in mitochondria. It has been suggested that FDXR is a putative contributor to TP53-mediated apoptosis from 5-FU in tumor cells through the generation of oxidative stress in mitochondria.37, 38 It has been shown that FDXR gene expression levels are higher in tumor than in normal tissues from colorectal cancer patients.39 In metastatic colorectal cancer patients treated with 5-FU plus leucovorin, FDXR gene expression was higher in responding tumors compared with nonresponding ones.40 Our observation, which shows a DFS advantage for patients whose tumors have low-FDXR mRNA expression compared with patients with high-FDXR tumor expression, is in contrast with the latter findings. However, a direct comparison between the results of these two studies is difficult as they were obtained in different disease stages (i.e., Stage IV patients in the study of Ichikawa et al.40 and Stages II-III patients in our study).

When patients are analyzed according to their disease stage, a significant inverse relationship is observed between DUT or NM23-H1 and DFS. Low-expression levels of these genes are significantly associated with a prolonged DFS in Stage III patients. Similar associations were observed for TYMS and RRM1; however, in these cases only a statistical trend was obtained.

NM23 is a putative metastatic suppressor gene. We studied two homologues, NM23-H1 and NM23-H2 (this last showed no relationship with DFS) that encode two polypeptide subunits of a nucleoside diphosphate (NDP) kinase. To date, the role of NM23 has been investigated in colorectal cancer only as a potential prognostic factor.41–45 However, the results of these studies are conflicting. Both an association between low expression of NM23-H1 and shorter survival43 and an association between low NM23-H1 expression and improved survival44, 45 have been reported. In other studies, no such correlations have been established.41, 42

In our experience, patients with low NM23-H1 gene expression levels benefited from adjuvant 5-FU chemotherapy more than patients with high levels. Based on the role of NDP kinase in the 5-FU activating pathway, it is conceivable that patients with higher enzyme expression in their tumor would experience a greater benefit from 5-FU therapy. Further studies will be needed to clarify this association.

TS, which is codified by TYMS, is believed to be the main determinant of 5-FU response.2 Although TS has been the most studied 5-FU-related biomarker in colorectal cancer patients, its predictive value in the outcome of 5-FU-treated colorectal cancer patients remains to be established.15 Johnston et al.46 first demonstrated a correlation between low-TS levels and improved 5-year DFS and OS in rectal cancer patients receiving 5-FU adjuvant chemotherapy. A meta-analysis by Popat et al.15 showed that colorectal cancer patients with advanced disease treated with TS inhibitors had a significantly better OS if they had low-TS expression in primary tumor or metastases, whereas a similar predictive role of TS expression was not established for the adjuvant setting. The lack of predictivity of TS expression in the adjuvant setting might be due to the fact that patients with high-expression levels seem to benefit from adjuvant 5-FU-based chemotherapy more than patients with low-TS expression, thus improving their DFS or OS.46–49 Of importance is the recent observation that TS status/expression does not consistently explain the variance of response of fluoropyrimidine-treated colon cancer cells in part due to RNA-based toxicity.36

A tendency of Stage III colorectal cancer patients whose tumors expressed low levels of TYMS to have prolonged DFS was observed in our study as well as in other previous ones15 and requires further investigation in wider patient populations.

Finally, ribonucleotide reductase (RR), which consists of two subunits, M1 and M2 (RRM1 and RRM2, respectively), catalyzes de novo synthesis of deoxyribonucleoside diphosphates as building blocks of DNA. In the 5-FU pathway, RR may convert fluorouridine diphosphate to fluorodeoxyuridine diphosphate that will be successively converted to fluorodeoxyuridine triphosphate and incorporated in DNA. Our results show a trend for increased RRM1 mRNA expression in patients with shorter DFS. There are very few reports on the expression50 or activity50, 51 of RRM1 and 5-FU sensitivity in colorectal cancer. Although reduced activity of RRM1 in a human colorectal cancer xenograft 5-FU-resistant compared with a 5-FU-sensitive one was observed, no difference in the expression of the RRM1 mRNA between the two xenografts was found.50 Similar results have been observed in colorectal cancer patients where a low activity of RRM1 was associated with 5-FU resistance.51 However, in this case, protein RRM1 levels were evaluated, thus preventing a direct comparison with our results.

We did not observe any relationships between the mRNA expression of other genes frequently studied in clinical samples for their involvement in 5-FU resistance (e.g., DPD, TP, OPRT, UMPK17, 21, 22 and therapeutic outcome in our case series. The correlations reported in the literature were obtained in metastatic colorectal cancer rather than in the adjuvant setting.

Despite the limited number of patients, these results suggest that the expression of specific genes involved in 5-FU metabolic pathway may be a potentially useful marker for predicting clinical response to 5-FU adjuvant chemotherapy in colorectal cancer. Data from this and related studies may help to identify gene profiles useful in selecting patients eligible to receive personalized adjuvant fluoropyrimidine-based therapy. Validation in an independent case series is warranted.